The 5 questions to test the validity of a statistical analysis

In this instalment of Sunday Skills, I want to direct you to a podcast from Stephanie Thomas at The Proactive Employer.

What to do when your results are statistically (in)significant lasts about 15 minutes and you won’t regret spending this quarter of an hour listening to Stephanie expand on the 5 questions you should ask yourself when interpreting the results of a statistical analysis.

  1. How would the results change if we had more observations ? Can you really draw conclusions or is the sample too small ?
  2. How well does the model fit ? Are we studying the appropriate relationships between the variables ?
  3. Are the results reasonable ? Use your professional experience and judgement to assess of the results make sense or not.
  4. Is the estimated disparity big enough to matter ? Consider practical significance independently of statistical significance. This really matters in countries where some categories of population are protected, for example females, ethnic minorities, people of certain ages etc.
  5. What follow-up is required ?

Stephanie includes examples from compensation analysis and clearly explains how the answers to these 5 questions will help you make decisions (or not) based on the results of the analysis you just performed on your compensation data.

I hope you enjoy this audio format for expanding your analytical skills !


Related posts :

Print Friendly, PDF & Email

Speak Your Mind